10 questions about product recommendations

Many retailers use automatic product recommendation technology to increase their sales and conversion rates. These recommended items are generally dynamically generated on commercial sites and are generally based on the purchase habits of specific customers or a group of customers.

Strands recommender is a leading provider of dynamically generated product recommendations. We recently interviewed its marketing manager Trevor legwinski about the concept of product recommendations and its effect on e-retailers.

10 questions about product recommendations

Practical ecommerce:What are product recommendations? How do they work on a retailer's website?

Trevor legwinski:"Because e-commerce companies have a lot of products and receive the optimal conversion rate, it is very difficult to increase sales and cross-selling in a structured manner. We recommend that you learn the personal and overall behavior levels of your website visitors to solve this problem. We automatically collect this data and commercialize your site to match the behavior and intent of new visitors with thousands of converted shoppers. For our return visitors, our products based on their customer information are the preferences that our engines learn anonymously over time, including preferences, favorite brands, or even a specific color and size."

PEC:Where do recommendations usually appear on the site?

Legwinski:"Retailers usually put recommendations on their home pages, product details pages, and their shopping cart. In addition, we have introduced recommendations on the classification page and search page, and even created a customer "personalized login page" similar to amazon.com "."

PEC:How do you decide to recommend those products?

Legwinski:"If your company is a customer, our engine anonymously detects that the customer accessing your site creates a customer profile based on his/her shopping behavior. These include, for example, products they browse, searches, add to shopping cart and permanent purchases. For new visitors, we show the items that people with similar behavior in the past are most likely to buy. For those who return, we visit their introduction and present products related to their preferences. Furthermore, we provide easy-to-use commercial tools that allow retailers to filter recommendation results based on their own market and business needs ."

PEC:What kind of results do retailers use for product recommendations? Can you reference some actual data from your customers?

Legwinski:"We work with retailers in almost every catalog, from clothing to footwear to electrical books, games, music, furniture, and sports equipment. On average, they can see an increase of 8 to 12 age points. Other factors that affect the performance of the recommender are the traffic magnitude, number of Recommendation sections, recommended placement location, and directory size. Give you some details and figures: an outdoor device customer uses our $349 standard monthly plan, with an annual revenue of about $1 million, currently, by using our services, we also get sales of $2000 per month. One of our largest apparel retailers has used our business program and has already received a return of 200 million and is receiving an additional million monthly return ."

PEC:The goal is to increase the basic conversion rate. Is there an average order size?

Legwinski:"Our recommendation engine aims to increase the average order size and conversion rate. By providing personalized experience for new visitors and visitors, we can increase the number of products added to the shopping cart while increasing the conversion rate of our customers. Also, using personalized emails can increase the response rate of existing customers ."

PEC:Is your recommendation service deployed in your company? How does it integrate with a retailer's website?

Legwinski:"Strands recommender is an SaaS solution, meaning it is completely deployed by us. This eliminates many version control issues and deployment complexity. Integration only requires three steps:

1. "Retailers upload their item directories to our systems through a file or link. You only need to perform this operation once because the directory is automatically refreshed. The format is very similar to the product feed you will send to a comparison shopping engine. This feed helps our engine learn your product types and maintain a Real-Time Correspondence with your directories. We also support data feeds from rankings, comment providers, and site search providers.
2. "The second step is user behavior on your site, which is then used to anonymously build a profile for each visitor. We are able to use such as "the person who purchased the product also bought ......" Algorithms detect shopping patterns, related profiles, and other more complex patterns. To do this, the simple tracking code similar to Google statistics is pasted to the bottom of your home page template, product details, shopping cart, order confirmation, and available "Desire list" and "favorite" pages.
3. "The last step is to present the recommendations to your viewers. Our customers can quickly copy and paste from a predefined recommendation pendant to any page on the e-commerce website. These pendants are based on best practices and usually work well for most of our customers. The system also provides a pendant editor that allows you to customize the appearance, feel, and content of each pendant. In this way, we make sure that each pendant fully matches your site and email template ."

PEC:Does your service support all shopping carts? Are you locally deployed or certified?

Legwinski: "Yes. Our JavaScript and API installation makes it very easy for retailers that install any shopping cart with our software. (Our JavaScript or API install makes it very easy for merchants with any shopping cart to install our software.) We have further simplified the creation of a shopping cart plug-in for Miva merchant and magento stores.

"We recently opened our API for developers to allow them to create shopping cart plug-ins for their favorite shopping cart and create interesting features on the user profile and collect Product Behavior of the strands recommendation system. We are very excited to see the developer group embracing personalization and creating some useful and cool plug-ins ."

PEC:How much does it cost?

Legwinski:"Our plan starts from $149 per month for small retailers to $999, which is higher for large enterprise customers. Recommendation systems are becoming more and more affordable. In the past, this technology was affordable to large retailers for thousands of yuan a month. Our goal has made our technology affordable for small retailers ."

PEC:Does a retailer have to have a certain size or a certain number of SKUs?

Legwinski:"We have found that retailers have at least 500 products and at least independent visitors each month will get the maximum profit. This does not mean that only small directories cannot benefit from recommendations, but the key to success is the traffic level and product mix. The receng requires a certain level of product data and effective traffic. We also found that independent visitors are the lowest threshold for continuous impact ."

PEC:What else do retailers need to know about product recommendations?

Legwinski:"In the past, email marketing was a positive recommendation that we did not emphasize. As a regular online shopper, I receive marketing emails from different retailers every day. But I'm frustrated that many retailers still send a standard template containing the same item to every customer. There are some ideas to give personalized emails to those customers who spend their time presenting product subsets or promotions that they may be concerned about. It tells customers that you are actually trying to serve them, and more, they will buy more from your extra efforts. "

-----------------------------------------------------------------------------

Note: This note is displayed in the first chapter of "mahout in action. Interesting. Translate it. The more translation, the more I think it is the marketing advertisement of this recommendation engine. However, the translation is still complete. The level is rough. The word profile seems difficult to express in Chinese.

Original article: http://www.practicalecommerce.com/articles/1942-10-Questions-on-Product-Recommendations